High Compression Ratio Turbo Gasoline Engine Operation
Using Alcohol Enhancement
by
Raymond Lewis
B.S. Mechanical Engineering
Massachusetts Institute of Technology, 2011
Submitted to the Department of Mechanical Engineering in Partial Fulfillment of the
Requirements for the Degree of
MASTER OF SCIENCE IN MECHANICAL ENGINEERING
AT THE
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
SEPTEMBER 2013
0 2013 Massachusetts Institute of Technology.
All rights reserved.
ARHiVES
MASSACHUSITS
OF TECHNOLOGY
NOV 12 2013
L-IBRARIES
Signature of Author ...................................
.
. .. . ........ .....
Department of Mechanical Engineering
August 19, 2013
Certified by.......................
John B. Heywood
Sun Jae Professor of Mechanical Engineering, Emeritus
Thesis Supervisor
Accepted by......................
David E. Hardt
Ralph E. & Eloise F. Cross Professorship of Mechanical Engineering
Chairman, Department Committee on Graduate Students
E
High Compression Ratio Turbo Gasoline Engine Operation Using
Alcohol Enhancement
by Raymond Lewis
Submitted on August 19, 2013
Submitted to the Department of Mechanical Engineering in Partial Fulfillment of the
Requirements for the Degree of Master of Science in Mechanical Engineering
Abstract
Gasoline - ethanol blends were explored as a strategy to mitigate engine knock, a phenomena in
spark ignition engine combustion when a portion of the end gas is compressed to the point of
spontaneous auto-ignition. This auto-ignition is dangerous to the operation of an internal
combustion engine, as it can severely damage engine components. As engine designers are
trying to improve the efficiency of the internal combustion engine, engine knock is a key limiting
factor in engine design.
Two methods have been used to limit engine knock that will be considered here; retarding the
spark timing and addition of additives to reduce the tendency of the fuel mixture to knock. Both
have drawbacks. Retarding spark reduces the engine efficiency and additives typically lower the
heating value of the fuel, requiring more fuel for a given operating point.
To study this problem a turbocharged engine was tested with a variety of combinations of
gasoline and ethanol, an additive with very good anti-knock abilities. Pressure was recorded and
GT Power simulations were used to determine the temperature within the cylinder. An effective
octane number was calculated to measure the ability of the fuel to resist knock. Effective octane
numbers varied from 91 for UTG91 to 111 for E25, respectively.
Engine simulations were used to extrapolate to points that couldn't be tested in the experimental
setup and generate performance maps which could be used to predict how the engine would act
inside of a vehicle. It was found that increasing the compression ratio from 9.2 to 13.5 leads to a
7% relative increase in part load efficiency. When applied in a vehicle this leads to a 2-6%
increase in miles per gallon of gasoline consumption depending on the drive cycle used. Miles
per gallon of ethanol used were significantly higher than gasoline; 141 miles per gallon of
ethanol was the lowest mileage over all cycles studied.
Thesis Supervisor: John B. Heywood
Title: Sun Jae Professor of Mechanical Engineering, Emeritus
2
(This page was intentionally left blank)
3
Table of Contents
Abstract ......................................................................................................................................................... 2
Abbreviations ................................................................................................................................................ 6
1.
Introduction .......................................................................................................................................... 8
1. 1.
Ethanol Blends .................................................................................................................................. 9
1.2.
Spark Retard ...................................................................................................................................... 9
2.
Experimental M ethods ........................................................................................................................ 10
Defining Knock Onset ...................................................................................................................... 10
2.1.
3.
Sim ulation Tools .................................................................................................................................. 11
GT Power ......................................................................................................................................... 11
3.1.
3.1.1.
Burn Rate .................................................................................................................................... 12
3.1.2.
Variables in the Sim ulation ......................................................................................................... 13
3.1.3.
Com parison of GT Power with Experimental Results ................................................................. 13
3.1.4.
Fuel Com parison ......................................................................................................................... 21
4.
Evaluating Anti-Knock Properties - The Knock Integral Approach ..................................................... 23
5.
Additions to the Perform ance M ap .................................................................................................... 28
5.1.
Extrapolation of Burn Rates ............................................................................................................ 28
5.2.
M BT ................................................................................................................................................. 29
5.3.
W ith Spark Retard ........................................................................................................................... 33
6.
Efficiency and Com pression Ratio ....................................................................................................... 37
7.
Autonom ie .......................................................................................................................................... 40
7.1.
Base Vehicle .................................................................................................................................... 43
7.2.
Perform ance in the Driving Cycle ................................................................................................... 43
7.2.1.
Fuel Economy .............................................................................................................................. 43
7.2.2.
Ethanol Percentages ................................................................................................................... 46
7.2.3.
Efficiency ..................................................................................................................................... 47
8.
8.1.
9.
Com parison with Experim ental Data .................................................................................................. 48
Experim ental Knock Onset at Different Speeds .............................................................................. 50
Revised Results ................................................................................................................................... 53
9.1.
Ethanol Fraction .............................................................................................................................. 55
10.
Findings and Conclusions ................................................................................................................ 58
10.1.
GT Power M odeling .................................................................................................................... 58
4
10.2.
Com parison of Different Fuels .................................................................................................... 58
10.3.
Octane Num ber ........................................................................................................................... 58
10.4.
Higher Com pression Ratio .......................................................................................................... 58
10.5.
Spark Retard ................................................................................................................................ 59
10.6.
Autonom ie Results ...................................................................................................................... 59
10.7.
Dependence of Knock Onset on Speed ....................................................................................... 59
10.8.
Future W ork ................................................................................................................................ 60
References .................................................................................................................................................. 61
5
Acknowledgements
First, I would like to thank Professor Heywood for taking me on as a graduate student on this
project, and for giving me guidance and support throughout the course of this project. His
knowledge, patience and advice helped me a great deal as I worked on this project.
Next I would like to thank the administrative staff, both at the Sloan Automotive Laboratory and
at the Mechanical Engineering Graduate Office for helping to ensure the work went as smoothly
as possible.
Finally I would like to thank friends and family for their continuing encouragement and
understanding.
6
Abbreviations
BMEP: brake mean effective pressure
CR: Compression Ratio
E10: 10% ethanol and 90% gasoline by volume
E20: 20% ethanol and 80% gasoline by volume
E25: 25% ethanol and 75% gasoline by volume
E50: 50% ethanol and 50% gasoline by volume
E85: 85% ethanol and 15% gasoline by volume
HWFET: Highway Fuel Economy Test Driving Cycle
MBT: maximum brake torque
RON: Research Octane Number
UDDS: Urban Dynamometer Driving Cycle
US06: supplemental driving cycle
UTG91: regular test hydrocarbon gasoline, 91 RON octane
UTG96: premium test hydrocarbon gasoline, 96 RON octane
7
1. Introduction
A recent trend in internal combustion engine design has been towards increasing the efficiency
of the engine. This is largely due to the fact that the price of gas is extremely variable and
mileage, a measure of how much gasoline is consumed by the engine is often a selling point used
to market vehicles. Strategies used to improve the efficiency of internal combustion engines,
include increasing the compression ratio and using a turbocharger.
Both strategies can be used to increase the efficiency in different ways. Increasing the
compression ratio increases the pressure inside the cylinder, which increases the work that can be
done over a given cycle with the same initial conditions. A turbocharger compresses the air
inside the cylinder, allowing for more air to be fed in. This leads to an increase in the available
output power, because more air means more fuel is available to combust. The relative magnitude
of efficiency loss mechanisms decrease with increased turbocharging, so as a result, the engine
becomes more efficient. [1]
While both approaches improve the efficiency of an internal combustion engine, an added
disadvantage in spark ignition engines is the increase in the tendency of an engine to undergo
engine knock. Engine knock is a phenomenon in an internal combustion engine, in which the
pressure and temperature of unburned gas reach a point where it spontaneously ignites.
Typical combustion in a spark ignition engine can be modeled as a flame front beginning at the
spark plug during spark ignition and continuing through the cylinder as more of the fuel-air
mixture is combusted. As the flame propagates the combustion products expand, compressing
the unburned gas. In some situations, the pressure and temperature of the unburned gas becomes
high enough to ignite on its own. This causes shock waves which can severely damage the
engine.
Knock is typically most prevalent when operating a high loads and low speeds. High loads,
measured by the applied torque of the engine, correspond to high cylinder pressures and
temperatures which increase the tendency of a fuel to knock. Lower speeds mean longer
combustion times, and therefore more time to knock for a given operating point.
Preventing and/or limiting engine knock is a key limiting factor in improving the efficiency of
internal combustion engines. Increasing the compression ratio of an engine will increase the
pressure inside of it, promoting knock. Likewise, turbocharging will also increase the pressure
inside of the cylinder, promoting knock.
How much an internal combustion engine performance can safely be improved is limited by how
much engine knock can occur in the cylinder. Therefore in order to achieve improvements in
efficiency using these methods, engine knock must be mitigated.
8
1.1.
Ethanol Blends
In order to limit the tendency of fuel to knock additives are typically added. Examples of fuels
that have been used to limit knock are ethanol, methanol and water. Adding such fuels
contributes two separate effects which can limit engine knock. First, these additives increase the
fuel chemical octane number. Second, the heat of vaporization of such fuels is much higher than
gasoline. As the fuel is injected as a liquid and vaporized within the cylinder a higher heat of
vaporization will lead to cooler air in the cylinder. As a result of this charge cooling effect, the
mixture in the cylinder will be cooler, which lower its tendency to knock.
Ethanol has received particular consideration because of its potential as a carbon neutral
alternative to gasoline, as well as its use as a knock resistant fuel. However, while it has been
demonstrated as an effective anti-knock fuel it has several drawbacks. Ethanol has a lower
heating value that is significantly lower than typical gasoline. As a result, more fuel is required
for the same energy output when using ethanol, and other alcoholic additives.
One possibility which will be explored here is that of using two tanks, one with gasoline and one
with a knock resistant additive, in this case, ethanol. Such a configuration could enable the
engine to run on typical gasoline when not in regions where there is a risk of knock and add
ethanol from an auxiliary tank when operating in regimes where there is a high risk of knock. As
the tendency of a certain fuel to knock is increased, so is the amount of ethanol added to it.
1.2.
Spark Retard
Another, more common method to reduce engine knock in the engine is with spark retard. The
goal behind spark retard is to delay the ignition at the spark plug so that the fuel does not have
enough time to reach auto-ignition. This is because spark retard decreases the pressures in a
given engine, and makes the peak pressure point within the cylinder occur later in the cycle.
Both these effects decrease the tendency for knock to occur at a certain condition.
Doing this however decreases the efficiency of the engine. The reason this occurs is because the
work produced by the gas in the cylinder is equal to the integral of pressure times the differential
change in volume:
W = f PdV
Retarding the spark makes peak pressure occur later in the combustion/expansion stroke and
decreases its magnitude, lowering the output power.
This leads to another tradeoff. Spark retard would allow higher loads with the same fuel,
decreasing the relative amount of knock resistant additives but at the same time would decrease
the efficiency, requiring more energy, and as a result, more fuel, altogether.
9
2. Experimental Methods
A 2.OL Turbocharged Ecotec Engine was used for experimental measurements. This engine had
a compression ratio of 9.2, a total displacement volume of 2 Liters and a twin-scroll
turbocharger. The experimental plan was to run the engine at speeds of 1500 rpm to 3000 rpm at
various loads. In this way a performance map could be generated by recording MBT at every
point in the operating range. At each point a spark sweep was performed, during which the spark
timing was advanced or retarded as much as possible, subject to performance constraints.
While running the experiments several limits were imposed on the range of loads applicable to
the engine. First, the maximum pressure within the cylinder was not allowed to surpass 100
bars. Any higher pressures, given the high loads the engine would be subject to, would risk
blowing the cylinder head gaskets. This was done by retarding the spark with respect to MBT.
While this strategy decreased the efficiency it also enabled the engine to be run at higher loads.
Second, the exhaust temperature was kept below 950 C. This was the limit of the temperature
range of the current turbocharger. When this constraint was reached, the spark was advanced.
Two types of gasoline were used as a baseline: a 91 RON octane, similar to normal gasoline and
a 96 RON 'premium' octane. To consider the effect of blending ethanol, the 91 RON octane
gasoline was mixed with ethanol. Blends of 91 RON were done because the goal of this research
was to quantify the antiknock benefits of ethanol, which would be more noticeable if compared
with a lower octane gasoline. Also, 91 RON is closer to the type of gasoline consumers
currently use. Mixtures were done on a by volume basis and will be referred to by the ethanol
percentages, by volume. Mixtures tested are EO, E10, E20, E25, E50 and E85.
2.1.
Defining Knock Onset
When running the engine, knock onset could be determined by the audible noise it makes but
when analyzing the data a more rigorous approach was taken by using the pressure trace.
Because knock manifests itself as high frequency pressure oscillations a filtered pressure trace
was used to determine the point at which knock occurs. The steps used to determine knock onset
are as follows.
Pressure data was taken in the cylinder at a sampling frequency of 100 kHz. A Fast Fourier
Transform was then performed in MATLAB for every cycle. This converts the pressure trace
from the time domain into the frequency domain. Second, anything corresponding to a
frequency below 2 kHz or above 50 kHz was removed. 50 kHz was chosen as an upper limit to
eliminate aliasing. 2 kHz was chosen as a lower limit to eliminate noise and any changes in the
cylinder not related to engine knock. This was essentially a band pass filter.
An inverse Fast Fourier Transform was applied to the filtered signal to return to the time domain
and the largest value of each cycle was recorded. This value was named the knock intensity for
10
each cycle. Knock onset was defined as the following two conditions met; the largest knock
intensity was between 2 bar and 4 bar and the percent of cycles with a knock intensity greater
than 1 bar (the knock frequency) was higher than 10%. An example of a cycle where knock
onset has occurred is shown in Figure 1.
80
L
-
eL
.L..
701
605040-
a30
20
10
nU
0
100
200
r
r
r
r
300
400
500
600
-
r
700
800
Time (Crank Angles)
Figure 1: A sample case of how knock is measured. Pictured here are a smoothed pressure
trace (blue) and the high pass filtered knocking data used to measure knock onset (green).
3. Simulation Tools
The experimental setup was only used to test the engine in a small component of the entire
performance range. In order to consider cases beyond what could be run in the experiment, and
to explore how the engine will behave in a vehicle, two simulation programs were used; GT
Power and Autonomie.
3.1.
GT Power
Most of the constraints mentioned here could be relaxed by the use of knock additives and
stronger engine components, and the engine operating range extends much farther than what is
tested. The engine tested can reach speeds below 1000 rpm and above 6000 rpm. This is beyond
the limits of our experimental setup so in order to obtain results for situations outside of our
experimental range, GT Power was used.
11
GT Power is a simulation package specifically designed for the study of internal combustion
engines. A model of the engine was provided by colleagues at GM (the manufacturer of the
engine we used) and used to get results for situations beyond our ability to run experimentally.
Using this simulation tool it was possible to create a performance map for the engine over the
entire operating range.
The performance map was expressed as a graph. Speed was on the x-axis, BMEP was on the yaxis and a contour of brake fuel conversion efficiency was superimposed on the graph. BMEP
was chosen because it was a measure of the work available from the engine that is independent
of the engine size.
BMEP=
n
2nTn
Here P is the power produced by the engine, N is the speed T is the torque, n is the number of
revolutions per cycle (2 in this case because this was done with a four stroke engine) and VD is
the displacement volume. The displacement volume is the volume the cylinders of the engine
sweep in one stroke; 2 Liters for this engine.
3.1.1. Burn Rate
The models used required the burn rate inside of the cylinder. This was calculated from
experimental data assuming that combustion in the cylinder could be modeled using two zone
poly-tropic processes; expansion of burned species and compression of unburned species.
1
Vu + Vb = VuO
V
()+
pA 1
P
= XbVf, Vuo = (1 - Xb)VO
Here xb is the fraction of fuel burned V is the (time dependent) volume of the cylinder P is the
cylinder pressure, Vu and Vb are the volume of the burned and unburned fuel mixtures, Po and Pf
are the initial and final pressures, Vo and Vf are the initial and final volumes and Vbf and Vuo are
intermediate variables. The poly-topic exponents, nu and nf, are calculated using the slope of
log-log fits of the pressure v volume curve before combustion begins, and after it finished
respectively.
Using these equations to solve for xb in terms of values which can be measured leads to the
equation below:
(V(P)n-VO(PO)
(nu
( 1
i f
nf
Pf)fu
=
Xb
(P)nf-VO(pO)n
12
This equation was solved for each case and fed into GT Power, which was then used to simulate
the performance of the engine. The initial and final states, denoted by the subscripts 0 and f are
the points of ignition, i.e. spark, and the opening of the exhaust valve.
3.1.2. Variables in the Simulation
The GT Power simulation allowed for a wide range of variables that can be adjusted to modify
conditions in the engine. In order to match simulations with results several variables were
adjusted. These included heat transfer to the cylinders, fraction of fuel burned during the cycle,
amount of charge cooling used to cool the fuel-air mixture etc.
Heat Transfer: GT Power has several built-in heat transfer models for calculating the heat
transfer from the fuel-air mixture to the cylinder walls. The models used typically underestimate
the heat transfer to the environment and therefore a scaling factor was included to adjust how
much heat transfer actually occurs. A scaling factor of 1.1 to 1.4 is typically recommended as a
beginning approach. After matching experiments to data it was determined that values between
1 and 1.2 are accurate.
Fraction of Fuel Burned: Combustion efficiency is not unity in the cylinder, and as a result, not
all of the fuel inside of the cylinder is ignited during the combustion phase. Typical values of
combustion efficiency for current engine range between 93% and 97%. Experimentally it was
found that 95% generates a good match between theory and experiment.
Charge Cooling: When injected into the cylinder, the fuel is a liquid which vaporizes in the
cylinder, cooling anything it comes in contact with. Ideally the fuel should only cool the mixture
but a fraction of the fuel will come in contact with, and by evaporation, cool the cylinder walls.
How much this occurs is beyond the scope of this study. To provide a bench mark, this question
was studied by colleagues at GM using a proprietary CFD model of engine evaporation in the
engine cylinder. This work was done as part of a previous study at base boost at 2000 rpm. [2]
They found that, irrespective of the fraction of ethanol within the cylinder, thirty percent of the
charge cooling effect available via the evaporating fuel was absorbed by the cylinder walls,
leaving seventy percent to cooling the mixture. This was kept consistent throughout the tests
conducted here.
3.1.3. Comparison of GT Power with Experimental Results
In order to validate the GT Power model's ability to simulate real world test conditions GT
Power was used to simulate conditions run in the experimental engine. The most extensively
tested fuels were gasoline (UTG96) and E85.
E85: The most extensively tested fuel was E85, chiefly because it yielded the best anti-knock
benefit of all the fuels tested. A performance map done using GT Power data was shown in
Figure 2. What can be seen here is that at low loads the efficiency increases quickly with load,
but at high loads it stagnates. This is due to the peak pressure constraints, seen in detail in Figure
13
3. Here peak pressure is plotted as a function of NIMEP. At the low load points, there is a
closely linear relationship between load and pressure, but at the high load points the pressure
constraint leads to a leveling off, as spark is retarded to avoid exceeding 100 bars.
220-
0
22 Contour
GT Power Results
Jo
200
1816
14(-
w
12
m
10;
8
8
1
74
1500
333
02
9
7
2000
2500
3000
Engine Speed (rpm)
Figure 2: GT Power simulation of the experimental performance map for E85.
14
120-
1000
UO)
L2
0
0
80S0')
C(
60-
070
0
Cl)
a_)
C
0
0~
00
6
0
40 -
1500
2000
2500
3000
RPM
RPM
RPM
RPM
07
20-
0
5
10
15
20
25
NIMEP (bar)
Figure 3: Peak Pressure at highest load timing for cases done using E85. A linear fit of peak
pressure v. NIMEP for data below the peak pressure limitation is shown, as is the peak
pressure limit. Note that peak pressure levels off as the peak pressure limitation is reached.
Comparisons of GT Power with the experiment at different speeds are shown in Figures 4-7.
The simulation and experiment are in close agreement here, which validates the model and
parameters chosen as accurate for simulating engine conditions.
15
>.w
34
32
-
-
30
.
28
0
26
24
C
GT Power
Experiment
0
o
22
W
20
LL 18
16
-
M
14
0
4
2
6
8
14
12
10
16
18
BMEP (bar)
Figure 4: Ethanol Brake Fuel Conversion Efficiencies at 1500 RPM
40
j35
C C30
-5
0
C)
U) 25
GT Power
Experiment
*
0
020
LL
Q)15
L-.
Mo 10
0
5
10
15
20
25
BMEP (bar)
Figure 5: Ethanol Brake Fuel Conversion Efficiencies at 2000 RPM
16
, 36
G3 4
32
-1
C 30
0
*
-
28
GT Power
Experiment
C 26
0
024
LL
-
C
22
I-
20
&
2
1O
m18 0r
10
5
20
15
25
BMEP (bar)
Figure 6: Ethanol Brake Fuel Conversion Efficiencies at 2500 RPM
>
36
S32
wS 30
---
L.28
C 26
0
*
GT Power
*
Experiment
24
_
22L.L
218
S18
- ------
-
0
5
10
---
15
20
25
BMEP (bar)
Figure 7: Ethanol Brake Fuel Conversion Efficiencies at 3000 RPM
17
UTG96: Premium gasoline was also tested at a wide enough range to generate a performance
map, shown in Figure 8. Being a gasoline it was more susceptible to knock than the higher
ethanol blends. As a result the highest loads tested here were much lower than initially used with
E85. A comparison of GT Power with experiment was shown in Figures 9-12.
16-
0
14
Contour
GT Power Results
-
12
.P 10
-3
cc
0 -.
.....
LjJ 8
W88
0-2-
30
m6
4
42
1&
1500
2000
2500
3000
Speed (RPM)
Figure 8: Turbocharged 2L DI Engine Performance Map Simulated in GT Power with UTG96
gasoline.
18
>% 35
30
w
C 25
0
GT Power
Experiment
20
0
15
10
C
5
-12
0
2
4
6
8
10
12
14
16
18
BMEP (bar)
Figure 9: UTG96 at 1500 RPM
>
35
C
aQ 30
.
25
C
-o 20
GT Power
Experiment
0
D15
C
0 010
S5
U0
C
i- -5 5
0
5
10
15
20
BMEP (bar)
Figure 10: UTG96 at 2000 RPM
19
40
35
S30
Li
C 25
0
L.20
C
0
15
GT Power
Experiment
- -
010
5
LL
U1)
0
-h
C
-5 '-2
0
2
4
6
8
10
12
14
16
BMEP (bar)
Figure 11: UTG96 at 2500 RPM
40
35
-.F 30
a)
25
U- 20
C
0
CLL
U)
15
GT Power
Experiment
10
76
5
UU)
Cu
0
-5
-10
-2
0
2
4
6
8
10
12
14
16
BMEP (bar)
Figure 12: UTG96 at 3000 RPM
20
3.1.4. Fuel Comparison
A comparison of brake fuel conversion efficiencies was done for the fuels tested at low loads to
determine how the engine behaves under MBT conditions. At high loads the fuel dependent
constraints due to knock lead to a deviation from MBT timing. As a result high load behavior
varies significantly from fuel to fuel. At low load fuel characteristics were expected to be
similar. Results of efficiency v. BMEP were shown in Figures 13-16.
>* 35
C~30 I-
0
-
I-
25
o 20
-
1,q
i
/
6
E85
UTG96
i
i
/
LIL015
-)
I
2
0
2
4
6
8
10
12
-I14
16
18
BMEP (bar)
Figure 13: Fuel Comparison at 1500 RPM
21
40
30
C 25
0
L.20
(D
C
E85
UTG96
15
0
0 10
LL
5
U)
0
-5 E
-5
m
_
_
r
0
5
F
10
20
15
25
BMEP (bar)
Figure 14: Fuel Comparison at 2000 RPM
40
U)35
:30
-
-
---
--- -
C 25
0
E 20
U)
C
0
E85
UTG96
15
010
_-
LL
-
_1_
_
_
_
0
Im
-5
-2
0
2
4
6
8
10
12
14
16
18
BMEP (bar)
Figure 15: Fuel Comparison at 2500 RPM
22
>%40
C)
c)35
/-
-
w0
o0
25
U)20
0
*
015
E85
UTG96
310
LL
mo)
r_
-5
0
10
5
15
20
25
BMEP (bar)
Figure 16: Fuel Comparison at 3000 RPM
These results show the brake fuel conversion efficiency as a function of load for the different
fuels. It was determined from these data that efficiency is essentially independent of fuel as a
first order approximation.
4. Evaluating Anti-Knock Properties - The Knock Integral Approach
In order to compare the effectiveness of each fuel at preventing knock an "effective" octane
number was developed using a model for Auto-ignition. It is assumed that the reaction can be
modeled as a one - step chemical reaction. In this case the reaction was modeled as a one step
Arrhenius reaction with a time constant r. At each point in the cycle the reaction can be said to
progress by an amount equal to:
Knock onset can be said to occur when the value of this integral reaches 1. The correlation used
in this investigation is as follows. [2]
T
ON3.402 p-1.7 exp
=17.68 (70-0)
ex(-800
_-_)
23
Here T is in milliseconds, ON is the effective octane number, P is the cylinder pressure in bars
and T is the temperature of the unburned mixture in Kelvin. By defining knock onset as the
point where the value of the integral reaches 1 an octane number can be found for each fuel.
Results for the UTG96 gasoline are shown below. Pressure and temperature are shown in Figure
17 and the knock integral evaluated at different trial octane numbers is shown in Figure 18. The
auto-ignition integral was a definite integral evaluated between two bounds, beginning of
compression and knock onset. Beginning of compression was set to the intake valve closure.
Knock onset was first assumed to occur at either peak pressure or peak temperature. Both were
tested here. More details are shown in the table. Shown are the crank angle and the octane
number where this induction/ auto-ignition time integral reached a value of one.
)00
80
Pressure (bar)
Temperature (K)
a:3
Cu
60
cc
a
8
-0
)0
40
a)
L_
CO)
U)
L_
-
20
0r
-160
-140
-120
-100
4
-80
-60
-40
-20
0
20
Sa)
C
200
40
CAD aTDC
Figure 17: In cylinder Unburned Mixture Temperature and Pressure plotted versus crank
angle degrees after TDC
24
2.5
ON=97 .2,Ocr= 2 3 ,Pressure Based
ON= 94.5, cr=22,Temperature Based
2
1.5
1
0.5
r_________
-160
-140
-120
_________
-100
r _________
-80
-60
r_________ r
~
-40
-20
0
20
40
CAD aTDC
Figure 18: Auto-ignition Integral Value assuming that knock onset (i.e. the value of the autoignition integral approaching 1) occur at peak pressure and peak temperature
Table 1: Comparison of Different Auto-ignition Octane Numbers, found using peak pressure
and peak temperature as the points of knock onset
Maximum Pressure
Maximum Temperature
Fuel Spec Sheet
23
22
97.1
97.2
94.5
Maximum pressure was reached slightly later than the maximum unburned mixture temperature
so the estimated octane number using maximum pressure is slightly higher. It also is closer to
the value given on the fuel specification sheet, so maximum pressure was used for future results.
The calculations for Figures 19-20 and Table 2 were done for a mixture of 25% ethanol and 75%
low octane (RON 91) gasoline, by volume (E25). Using maximum pressure as the point of
knock onset reveals a research octane number in the engine of 111 for E25. This can be broken
down into several components; the chemical octane number, the evaporative octane number, and
the fuel sensitivity benefit of the engine.
25
Chemical octane number is the contribution due to the auto-ignition chemistry of the mixture.
The gasoline component of this fuel has a research octane number of 90.8, which with an ethanol
volume percentage of 25% leads to a chemical octane number of about 99.
Evaporative octane number is the contribution with direct fuel injection due to charge cooling
through fuel evaporation. When injected into the cylinder, the fuel evaporates, cooling the in cylinder mixture in the process, which lowers its tendency to auto-ignite. Estimates of this
number for gasoline and ethanol are about 7 and 20, respectively. Since the fuel is 75% gasoline
and 25% ethanol, this contribution was estimated as 10.
100
)00 .
Pressure (bar)
Temperature (K)
E
a)
-__5
50
EL
)0
U-160
-140
-120
-100
-80
-60
-40
-20
0
20
-
40
CAD aTDC
Figure 19: In Cylinder Unburned mixture temperature and pressure for an E25 knocking case.
26
3
Li
---
-i
_
_
L _
__L
__L
I_
_
ON=106,0cr= 9 ,Temperature Based
2.5
0)
_
ON= 111, 0 cr= 1 0 ,Pressure Based
2
1.5
L..
0)
C
4.J
1
0.5
-1
-160
-140
-120
-100
-80
-60
-40
-20
0
20
40
CAD aTDC
Figure 20: Knock Integral Evaluation for an E25 knocking case.
Table 2: Similar approach to table 1 done using E25
Maximum Pressure
10
Maximum Temperature
111
9
106
Additionally, the sensitivity benefit is the increase in octane number due to differences between
the current engine and the standard octane test engine. This contribution amounts to about two
octane numbers, which in this case lead to a total octane number for E25 of:
99+10+2=111
This is close to the octane number (111) above, calculated using the auto-ignition time integral,
again showing the approach works in calculating the octane number of new fuels.
The integral was evaluated over an interval beginning at 144 crank angle degrees before top dead
center. This point represents the closing of the intake valve, when compression begins in
earnest. The results for different fuels are shown below. The numbers calculated using the data
from the GT Power model of the engine, as can be seen here are in most cases higher than the
27
references. As mentioned before, this can be due to uncertainties in the charge cooling effect and
the octane sensitivity. For fuels with ethanol in them, the possible variation in octane number
due to evaporative cooling was included as well.
Table 3: Octane Numbers done for additional fuels. Uncertainties due to charge cooling are
mentioned as well.
UTG91
UTG96
E10
E20
E25
Experimental Octane
Number
91
97
100
105
111
Chemical Octane
Number [3]
90.5
96.4
95.05
98.84
100.4
Variations due to Evaporative
Octane Number (if appropriate)
8.3
9.6
10
5. Additions to the Performance Map
GT Power was used to determine how much ethanol would be required to suppress knock at each
point by using the knock integral approach at each speed tested. This was done by using a PID
controller to force the value of the auto-ignition integral to reach one by adjusting the throttle and
waste-gate accordingly. To generate an upper bound on the knock onset lines the bounds of the
integral were set to intake valve closure and peak pressure. For each fuel tested a knock onset
line was determined from GT Power and superimposed on the performance map.
5.1.
Extrapolation of Burn Rates
Since the burn rate was set in GT Power as an input it was needed to extrapolate based on current
data taken from experiments done inside the engine. Data were taken with E85 at the range of
tested speeds, from 1500 RPM to 3000 RPM. Burn duration data, defined by the number of
crank angle degrees between 10% of the fuel burned to 90% of the fuel burned were shown in
Figure 21.
What can be seen here is that at low loads the burn duration in crank angle degrees decreases as
load increases and increases as speed increases. The sudden change in the trend above 1.4 bar
manifold air pressure is due to the fact that at higher loads the limitation of peak pressure leads to
a change in the behavior of the engine. Spark retard, which is implemented at high loads to limit
knock, leads to slower burn duration as a result of the lower pressures and temperatures in the
cylinder during combustion.
In order to deal with non-MBT behavior combustion retard was used to separate the cycles
operating within MBT from others. Combustion retard was found by taking the fifty percent
burned fuel crank angle degree of each point and comparing it with the point corresponding to
the highest load.
28
Combustion Retard =
0
OMBT -
case
If MBT was reached there should be positive and negative values of combustion retard.
Therefore, when extrapolating only those cases where this case is satisfied were considered.
32-
o
o
30 -
0
28
1500
2000
2500
3000
RPM
RPM
RPM
RPM
0
0)
0)
0 260
0 24-
0
0
0
0I
0
0
0
22
0
0
0
0
0
20-
0
0
0
0
c
18
0
16 0.2
r
r
r
0.4
0.6
0.8
r
r
1
1.2
1.4
1.6
r
r
1.8
2
MAP (bar)
Figure 21: Burn Rate data used to extrapolate for GT Power simulations. Only data below 1.3
bars were used in the extrapolation function, since the 'kink' indicates a change in behavior.
5.2.
MBT
Using GT Power performance maps of the engine were generated when operated at
different compression ratios. They were shown as contour plots of efficiency superimposed on
graphs of engine load, BMEP in bars v engine speed in RPM, as before, in Figures 22-24. The
knock onset lines were added onto the performance maps for cases run with gasoline (UTG91),
E10, E25 and E50.
29
- -
L
L
L
L
L
L
L
30
25
.0
a-
Efficiency
0 GT Power
___WOT
UTG91
E10
E25
20
0 __0 0
15
0
0-
-or
10
5
500
---20
16- -r) 12
r
)>(r
l
r
1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000
Speed (RPM)
Figure 22: Performance Map for a compression ratio of 9.2
:i
L
L
6.
L
6.
6
6
Efficiency
0
WOT
UTG91
E1 0
E2 5
0
30 H
d
25
20
0o0
U
-M
a-
GT Power
15
--90
-250
10
-
60
-0-2
5
500
1000
1500 2000 2500
3000 3500
4000 4500
5000
5500
6000
Speed (RPM)
Figure 23: Performance Map for Compression Ratio of 11.5
30
-
. ..........
- -------------
-----------..............
....................
------............
L
L
L
6
L
L
30 H
25
,
--
L
Efficiency
GT Power
WOT
UTG91
El 0
E 25
--
o
E5 0
20
L.
Cu
I0
15
10
5
6
-- -1--G- 47( 2~
-
500
1000 1500 2000 2500
-
21
-
G
3500
4000 4500
3000
.
21
(43
5000 5500 6000
Speed (RPM)
Figure 24: Performance Map for Compression Ratio of 13.5
Two conclusions can be reached upon looking at these data. First, efficiency increases when
increasing compression ratio. Second, as compression ratio increases, the knocking threshold for
each fuel decreases. Both trends are due to an increase in the pressure of the cylinder due to the
higher compression in each cylinder. To illustrate this, peak pressure data for the three
compression ratios is shown in Figure 25-28. For the compression ratio of 9.2 the peak pressure
surpassed 100 bars at about 18 to 20 bar BMEP, which in the experimental E85 performance
map was where efficiency began to level off. As compression ratio increases so does peak
pressure for a given load, increasing the work done by the cycle and increasing its tendency to
knock.
31
L
L
L
L
L
L
L
L
L
L
32
30 F
28
26
C
24
1 000
22
m20
0
020
0
18
00
16
00
14
12
r
500
1000
r
r
r
1500 2000 2500
r
r
3000 3500
r
r
4000 4500
r
5000 5500
6000
Speed (RPM)
Figure 25: Performance Map with Peak Pressure Contours at CR=9.2
L
L
L
L
L
L
L
L
L
.
30
00
25
1400
-
0ij
0
G
0
0
13
20
200
000
0
15
0
10
r
500
0
0
0
0
0
0
0
rl
r
r
r
r
r
r
1000 1500 2000 2500 3000 3500 4000 4500
0
r
5000 5500 6000
Speed (RPM)
Figure 26: Performance Map with Peak Pressure Contours at CR=11.5
32
L
L
LiL-
30-
25-
205--
170-160
-
01500
40
W20
w-
10
0
-
0
-1202
10--
15
0 0
/-
10
500
0
b-2-
0
0
0
0
0
1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000
Speed (RPM)
Figure 27: Performance Map with Peak Pressure contours at CR=13.5
5.3.
With Spark Retard
To consider the effect of spark retard additional maps were generated using the following
scheme. In regions where knock is not an issue the engine was run at MBT timing with UTG91.
When the engine begins to knock with the base gasoline, the spark was retarded, up to a certain
limit. The limits chosen here were 5 crank angle degrees after MBT timing and 10 crank angle
degrees after MBT timing. Once that limit was reached, then ethanol was added to reduce knock
while spark was still retarded at the limit.
Two trends can be seen when looking at these data in comparison with the data shown at MBT.
First, the knock onset lines were higher. Second, the efficiencies are lower at high loads. The
efficiency drops when retarding spark by 5 degrees and 10 degrees after MBT timing are on
average 2.05 and 4.22 respectively. Both are due to the lower pressures and temperatures in
retarded combustion. Lower pressures reduce the tendency of a fuel to knock, but they also
decrease the efficiency of the fuel. This is because, for essentially the same inlet conditions, a
lower pressure leads to less work, decreasing the efficiency.
33
C~
GT Power
WOT
UTG91
UTG91 5 deg ret
ELi 5 de ret
E25 5 de ret
E50 5 de ret
30!
-------
4b
25 F
-o
20
w
m
15-
34 -G
10
-
30~
-2
5
-4-
iPX
500
20
16
1000 1500 2000 2500 3000 3500 4000 4500
5000 5500 6000
Speed (RPM)
Figure 28: Performance Map with up to 5 degree retard at CR=9.2
L
L
L
L
L
30 F-
25
L
GT Power
WOT
UTG91
UTG91 10 deg ret
E10 10 deg ret
- E25 10 deg ret
cc
-o
20
U
w
15
10
5
500
-14-
-20
-----r 2r
16r
r
16r -2 - 121000 1500 2000 2500 3000 3500 4000 4500 5000 5500
r
6000
Speed (RPM)
Figure 29: Performance Map with up to 10 degree retard at CR=9.2
34
L
L
L
L
L
(? &GT Power
WOT
UTG91
UTG91 5 deg ret
E10 5 deg ret
E25 5 de g ret
E50 5 de g ret
L
30
25
-o 20
a15
--35-21
10
5
31-
-----
1.
r
21
17
r
500
21
-17
r
13
F
1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000
Speed (RPM)
Figure 30: Performance Map with up to 5 degrees retard with CR=11.5
GT Power
WOT
UTG91
UTG91 10 deg ret
(>
L
L
L
L
L
30-
g ret
E25 10 dEg ret
25-
E50 10 dEg ret
Ca 20Mo
a-j
w
M
15
1
0-5
-
125-
21
-
-
21
-17
13
0r
1000 1500 2000 2500 3000 3500 4 000 4500 5000 5500 6000
21
500
-252
9
52
--
7-
Speed (RPM
Figure 31: Performance Map with up to 10 degree retard at CR=11.5
35
L
L
GT Power
WOT
UTG91
UTG91 5 deg ret
E10 5 de g ret
E25 5 de g ret
g ret
L
30
-
25
SE50
-0 20
a-
15
10
5
500
-
25
----
5-
21
1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6C00
Speed (RPM)
Figure 32: Performance Map with up to 5 degree retard done at CR=13.5
30 -
GT Power
WOT
UTG91
UTG91 10 deg ret
25
E10 10 de g ret
E25 10 de g ret
E50 10 del gret
cc
-o
a
w
20
77
15
10
5
2
~-
31
___29i&
--
500
1000
9
-22
1
j7
13
1500 2000 2500 3000 3500 4000 4500 5000 5500 6000
Speed (RPM)
Figure 33: Performance Map with up to 10 degree retard at CR=13.5
36
5 de
6. Efficiency and Compression Ratio
Compression Ratio increased with efficiency and in order to validate the efficiency changes
noticed in these performance maps the efficiency was compared to that of other cases previously
studied. To compare compression ratios the efficiency was taken at the point corresponding to
one quarter times the maximum torque at 1500 RPM. This point was chosen because for the
majority of operation inside of a vehicle, the engine is near that operating point.
Net Indicated
40
038
tF=36
C
0
o
32 -
2
3
4
5
6
7
8
-
LL:3 30
28
8
9
10
11
12
13
14
Compression Ratio
Figure 34: Indicated Fuel Conversion Efficiencies compared to other cases published in the
literature
37
Brake
40
39
C) 38
37
W
C
36
35
0
a~)
--
LL
2
3
4
5
-
34
33
32
31
30
9
9.5
10
10.5
11
11.5
12
12.5
13
13.5
Compression Ratio
Figure 35: Brake Fuel Conversion Efficiencies compared to other data published in literature.
There was significant variation in the efficiencies shown because they were taken at different
points in the operating map. Some were taken at part load, where efficiencies are lower and
some are taken at wide open throttle, where efficiencies are higher. In order to generate a good
comparison the data were normalized by the efficiency at a compression ratio of 10:1. This
would normalize the changes in efficiency across different experimental runs. More information
about the cases beings shown was provide in Table 4. When changing compression ratio from
9.2 to 13.5, the compression ratios in the current study, part load efficiency went from 30.9 to
33.2, a relative increase of about seven percent.
38
Net Indicated
) 1.15
C
Q)
W
2
3
4
5
1.1
C
0
U)
( 1.05
8
0
1
-~
LL
_0
0
o
0.9
9
8
10
11
12
13
14
Compression Ratio
Figure 36: Normalized indicated fuel conversion efficiencies
Brake
1.14 r
1.12 1
0
C
0
I
-I
1.1
1.08
__
__
__
__
-
I
__
_
-2
3
4
5
I
1.06
1.04
1.02
--
1
F
U-
-o
0
I-
U) 0.98
N
0.96
0 0.94
z
9
9.5
10
10.5
11
11.5
12
12.5
13
13.5
Compression Ratio
Figure 37: Normalized brake fuel conversion efficiencies
39
Table 4: Information and references of the efficiency v compression ratio data
1
1500 RPM 7 bar MAP
1
Part Load
3
500 cc engine [6]
3
1500 RPM, 5 bar
BMEP [7]
5
6 bar IMEP, 2000 RPM [5]
5
2500 RPM, 5 bar
BMEP [71
7
2000 RPM
2 bar BMEP [81
7. Autonomie
To study how the engine would work inside of an actual vehicle, Autonomie was used to
simulate the engine when running different drive cycles. Autonomie is a vehicle simulation
program developed at Argonne National Laboratory. A map of fuel consumption as a function
of engine speed and torque was fed into Autonomie and, using said map, Autonomie could
simulate a vehicle when operating under different drive cycles. For every tenth of a second
interval the torque, speed and fuel consumption of the engine were recorded. This was used to
determine how the engine operates when placed inside an actual vehicle.
Three cycles were run in Autonomie: UDDS (meant to simulate city driving), HWFET (meant to
simulate highway driving) and US06 (meant to simulate high speed, high acceleration and high
variation in speed).
40
Figure 38: UDDS, meant to simulate highway driving [9]
Figure 39: HWFET, meant to simulate city driving [9]
41
Figure 40: US06 additional cycle [9]
Table 5: More Information about the Drive Cycles [11]
UDDS
HWFET
US06
20 mph
48 mph
48 mpl
17 miles
16 miles
13 miles
DescriptionI
Average Speed
Time
Distance
Max. Accel.
Since it has been determined that fuel conversion efficiency is independent of the fuel, the
performance maps developed in GT Power were used in Autonomie. The brake fuel conversion
efficiencies developed in GT Power were converted into fuel consumption rates, which were
used in Autonomie. It was assumed that the fuel flow rate at a given speed and torque were
expressed as follows:
f =N*T
Tj*LHV
Here f is the fuel flow rate in kg/s, N is the engine speed in rad/s, T is the engine torque in Nm, r
is the brake fuel conversion efficiency and LHV is the lower heating value of the reference fuel
in J/kg. Gasoline was chosen as the reference fuel. Using the efficiencies gathered from the
performance maps fuel flow rate was calculated at each torque and speed point and used as an
input to Autonomie. The output was defined as a gasoline equivalent, and by equating energy
consumption those results could be used to calculate gasoline and ethanol consumption.
42
7.1.
Base Vehicle
The vehicle chosen was a modified version of the conventional 2 wheel drive midsize sedan with
automatic transmission, based on the Toyota Camry. For all simulations reported here the engine
cylinder displacement was kept at 2.OL. Detailed information on the vehicle parameters is
shown in the table below.
Table 6: Vehicle Parameters used in the simulation [11]
Engine
Max Torque (Nm)
[441.8,
446.5,
448.01
Max Torque increased with
compression ratio
Gear Box
Gear Ratios
[2.563, 1.552,
1.022,0.727,0.521
5 speed gear ratio
Final Drive
Gear Ratio
4.438
Differential
7.2. Performance in the Driving Cycle
7.2.1. Fuel Economy
Using the data, three different metrics for fuel economy were developed. First, miles per gallon
of gasoline equivalent could be determined from the unprocessed Autonomie results, shown in
Table 7. This would serve as a proxy for the amount of fuel energy used over the course of each
driving cycle with each amount of allowable spark retard and at each compression ratio.
43
Table 7: Miles traveled per gallon of gasoline equivalent used.
UDDS
HWFET
US06
29.37
40.76
27.13
11.5 - MBT
30.10
41.95
28.33
-
30.51
41.91
28.15
30.42
41.56
28.60
9.2 - MBT
-
up to 5 deg retard
up to 10 deg retard
up to 5 deg retard
up to 10 deg retard
13.5 - MBT
-
up to 5 deg retard
up to 10 deg retard
From the data it can be seen that mileage increases with compression ratio. For the UDDS cycle
mileage increases from 29.3 to 29.93, a relative improvement of 2 percent. The HWFET cycle
mileage increases from 40.79 to 42.25 a relative improvement of 3.6 percent. The US06 driving
cycle leads to a relative increase of 5.7 percent; 28.81 to 27.24.
These increases were all lower than the 7 percent increase in brake fuel conversion efficiency
seen earlier. One reason for this difference is that the part load efficiency was taken at one
fourth times the maximum torque, a fixed point relative to the performance map. As the
compression ratio increased so did the maximum torque. As a result the part load point
increased as well. In the performance maps it can be seen that at low loads efficiency is highly
dependent on load, so this has the effect of decreasing the efficiency.
The fuel consumption over the entire cycle was converted to amount of gasoline and ethanol
usage. This was done in order to determine how much gasoline and how much ethanol would be
required for each case studied here.
Determining the amount of ethanol required was done by extrapolating the data used to create
the knock onset lines on the performance maps. As a result ethanol fraction could be expressed
as a function of speed and torque for each operating point. Since efficiency is independent of the
fuel it was assumed that the total fuel energy of the gasoline-ethanol blend is the same as the
gasoline equivalent. The resultant relationship is shown below:
rhsim * t * LHVref = Vg * PG * LHVgasoline + Ve * Pe * LHVethanol
44
Here libsim is the mass flow rate of the reference fuel from the simulation, t is the length of the
time-step, LHV denotes the lower heating value, V denotes the volume and p is the density. For
Autonomie t is 0.1 seconds. The subscripts gasoline and ref both refer to the gasoline used to
blend with ethanol, UTG91. The volume of ethanol and gasoline could be related to the ethanol
volume fraction Ef as follows:
(Ve + Vg) * Ef = Ve
Combining these two equations leads to the expressions below:
Vg =
!ilsim*t*LHVgasoline
Pg*LHVgasoline+
Ef
*Pe*LHVethano
VEf V
Ve e=g
1-Ef
Using these equations the volume of gasoline and ethanol could be determined at each point in
every cycle. Properties of the fuels relevant to these calculations are shown in Table 7.
Table 8: Fuel Properties Relevant to the Autonomie Based Calculations
Gasoline (UTG91)
Ethanol
Density (kg/m 3)
Lower Heating Value (MJ/kg)
789
29.69
Mileage was expressed below as miles per gallon of gasoline and miles per gallon of ethanol.
Two main results can be seen here. The first is that gasoline mileage can be higher than the
effective mileage. The largest different here is at the US06 cycle at MBT timing at a
compression ratio of 13.5. Here the miles per gallon of gasoline equivalent is 28.81 and the
miles per gallon of gasoline is 33.86; a relative increase of 17.5%. This is because the addition
of ethanol decreases the amount of gasoline required to provide the same energy; some of it was
provided by the ethanol.
Table 9: Miles traveled per gallon of gasoline consumed.
9.2 - MBT
- up to 5 deg retard
-
US06
29.36
40.76
27.36
30.24
42.14
30.43
30.51
41.91
28.57
30.75
41.84
30.46
up to 5 deg retard
- up to 10 deg retard
13.5 - MBT
- up to 5 deg retard
-
HWFET
up to 10 deg retard
11.5 - MBT
-
UDDS
up to 10 deg retard
45
The estimates of miles traveled per gallon of ethanol consumed were shown in Table 10. The
mileages shown here are significantly higher than that of gasoline. The lowest value of the entire
simulation was 141 miles per gallon of ethanol with the US06 cycle at MBT and a compression
ratio of 13.5. This is chiefly due to the fact that ethanol is only used where it is needed, when
there is a risk of knock. Knock is only required at high load points, where the efficiency of the
engine is higher than normal engine operation. For this reason the mileage estimates given for
ethanol are much higher than with gasoline, in most cases by several orders of magnitude.
Table 10: Miles traveled per gallon of ethanol used.
HWFET
US06
4829
6938
300
3.2*1
6.7*10
1398
2046
4575
341
UDDS
9.2 - MBT
-
up to 5 deg retard
up to 10 deg retard
11.5 - MBT
up to 5 deg retard
- up to 10 deg retard
13.5 - MBT
- up to 5 deg retard
-
-
-
up to 10 deg retard
7.2.2. Ethanol Percentages
Using the data shown above it was possible to calculate the percent by volume of ethanol
required under each driving cycle condition. Data were generated using the nine performance
maps and the three drive cycles. The results are shown in Table 11.
Table 11: Ethanol Percentage done using the performance maps
IJDDS
HWFET
US06
0
0
1.16
11.5 - MBT
0.62
0.6
9.2
-
0.0095
6*10-5
2
1.48
0.91
8.21
9.2 - MBT
-
up to 5 deg retard
up to 10 deg retard
up to 5 deg retard
up to 10 deg retard
13.5 - MBT
-
up to 5 deg retard
up to 10 deg retard
46
What can be seen here is that as the amount of allowable spark retard is increased the ethanol
fraction decreased. Going from MBT to allowing up to 5 degrees of retard has the effect of
reducing the amount of required ethanol by a factor of 2 to 6. This is due to the retarded
combustion leading to lower pressures and temperatures in the cylinder, decreasing its tendency
to knock. Also increasing compression ratio increases the amount of ethanol required by a factor
of 2 to 5. This is due to the increase in pressure and temperature caused by the higher
compression of the mixture.
7.2.3. Efficiency
In addition the average brake fuel conversion efficiency of each cycle could be determined by
taking the total work done by the engine and dividing by the total energy input.
ffmdt
f rLHVdt
Here 11 is the average brake fuel conversion efficiency, T is the instantaneous engine torque, o is
the instantaneous engine speed, LHV is the lower heating value of the fuel, th is the mass flow
rate and dt is the differential change in time.
Over the course of a driving cycle the engine can be used as a brake, when no fuel is fed into it.
In this case, the torque becomes negative and the friction inside the engine is used to slow the
vehicle down. This portion of the operation was not deemed relevant because no fuel is
consumed during this process and the energy lost to friction in the engine was supplied by the
fuel in the first place.
The results of removing the negative torque portion of engine operation were shown in Table 12.
A key result seen here is that increasing the compression ratio tends to increase the efficiency.
This is most pronounced in the US06 cycler, where at MBT the efficiency increases from 30.53
to 32.29. This leads to a relative increase of 5.5%. The smallest relative gain was about 2% for
the UDDS cycle. Another result seen with the US06 cycle is the decrease in efficiency when
spark retard in introduced. This is more pronounced with the US06 cycle because the US06
cycle includes higher loading points than the HWFET and UDDS cycle. As a result, the changes
associated with spark retard have a stronger effect on it.
47
Table 12: Efficiency calculated only considering points where the torque is positive
HWFET
UDDS
US06
26.13
23.54
30.41
26.88
24.13
31.76
9.2 - MBT
-
up to 5 deg retard
up to 10 deg retard
11.5 - MBT
-
up to 5 deg retard
up to 10 deg retard
268524.44
31.54
266324.37
32.05
13.5 - MBT
-
up to 5 deg retard
up to 10 deg retard
8. Comparison with Experimental Data
Estimated amounts of ethanol required for driving cycles were different for performance maps
generated by experiments and simulation. To compare the differences, wide open throttle lines
and knock onset lines of experiments and simulation were overlaid on the same graph. The first
major difference noticed was the shape of wide open throttle line. The wide open throttle line of
the simulation was higher than the experimental map since the data were limited by peak
pressure remaining below 100 bars in the experiments. This limitation was not applied in the
simulation. Also, they were different since the turbine behavior of the simulation was not exactly
the same as the actual turbine. The maximum manifold air pressure using simulation was 2.4
bars while it was 2.0 bars for the actual engine.
One thing that was noticed when looking at the results was that the knock onset lines predicted
by assuming knock occurs at peak pressure seem to be much more dependent on speed than the
experimental data previously recorded. In theory knocking behavior should be dependent on
speed because a lower speed would provide for more time during the combustion cycle for knock
to occur.
Differences between the simulation and experimental results are shown in the Figure 41 below.
Experimental results show knock onset to be much less dependent on load than what simulations
suggest. One interesting results shown here is that between 1500 rpm and 2000 rpm, the
simulation and experiment are in close agreement.
48
30.
25
20
15
10
WOT Simulation
WOT Experimental
Knock Onset Limit RON 91
Knock Onset Limit E10
5-
0
0
r
1000
r
2000
r
3000
r
4000
f'.
5000
6000
Engine Speed (rpm)
Figure 41: Comparison of simulation and experiment
One possible reason for the discrepancy was that experimentally knock onset can occur later than
peak pressure. The simulations assume knock occurs at peak pressure. That provides an upper
bound on knock onset lines for the majority of the map. In order to consider the possibility of
varying knock onset and to provide a range of possible knock onset constraints, knock onset lines
were generated for the same fuel (RON91) by widening the integration window used to predict
knock. The results of extending the integration window by 5, 10, and 15 crank angle degrees are
shown in the Figure 42 below.
Extending the integration window after peak pressure allowed more time for the end gases to
knock. Using the Douad-Eyzat correlation this corresponds to a higher integral value for all
cases being studied at the point define as knock onset. As the integral value becomes higher at
each operating point, the load point at which the integral value reaches I decreases.
Another trend that can be seen here is that as the window is extended further, the change in load
at knock onset decreases. The decrease in BMEP due to changing knock onset from peak
pressure to 5 crank angle degrees after peak pressure is larger than the change due to an
additional 5 crank angle degrees, almost by a factor of 2. This is because after peak pressure
both the pressure and the temperature decrease. As the integration window is widened the
decrease in BMEP becomes a diminishing trend.
49
16
Peak Pressure
5 degree after Peak Pressure
10 degree after Peak Pressure
15 degree after Peak Pressure
14
12
10
10
w
8
-
6
4F
1500
2000
2500
3000
Speed (RPM)
Figure 42: Here is the effect of changing the bounds of the auto-ignition integral to later than
peak pressure. Blue represents setting knock onset to occur at peak pressure. Green, red
and cyan are the results of setting knock onset to occur 5, 10 and 15 degrees after peak
pressure, respectively.
8.1.
Experimental Knock Onset at Different Speeds
In order to further explore the relationship between knock onset and speed data were taken at
conditions where knock onset occurred and studied. Sample data is shown. An attempt was
made to determine how peak pressure relates to knock onset by taking the filtered pressure trace,
originally used to determine knock onset and the pressure profile that result from subtracting
those results from the raw data. The shift in knock onset was determined by recording the
passage of time, in tens of microseconds from peak pressure between the two traces. The results
are shown for a set of example cases for UTG91 run at 1500 RPM, 2000 RPM, 2500 RPM and
3000 RPM. Each case is the worst knocking cycle at the spark timing and manifold air pressure
tested.
50
1500 RPM, delay = -40 microseconds
80
78 7674 -
CI
U)
KO
72 -
70
2..
CL
6866 -
Pmax
64 -
z
62 -
-I
60'
41 50
I I
4200
4250
4300
4350
Time (*10 microseconds)
Figure 43: Sample Knocking Data at 1500 RPM, 10ps=0.09 CA
2000 RPM, delay = 40 microseconds
80
L
L
L
L
78 76 -
KO
74
2CG70
13)
6866 Pmax
64 62-
60
r
3130
3140
r
r
3150
3160
r
3170
3180
3190
3200
Time (*10 microseconds)
Figure 44: Sample Knocking Data at 2000 RPM, 10ps=0.12 CA
51
2500 RPM, delay = 280 microseconds
80
L.
6
7876 -
KO
74 -
CU 72ID 70
~A~W<W~
U) 68
Pmax
66
64
{1
62
---
60
2400
2300
2500
2600
2700
2800
Time (*10 microseconds)
Figure 45: Sample Knocking Data at 2500 RPM, 10ps=0.15 CA
3000 RPM, delay = 310 microseconds
80
78 76 _--
KO
74
-o 72 FU)
70
U) 68 -
12-
66 -
Pmax
6462-
60
1800
1900
2000
2100
2200
2300
24
2400
Time (*10 microseconds)
Figure 46: Sample Data at 3000 RPM, 10ps=0.18 CA
52
These results show that knock onset can sometimes occur several crank angles later than peak
pressure. In order to get an average, data were analyzed by averaging over all knocking cycles.
A simple linear curve fit of the data is shown in Figure 47.
Knock Shifting for UTG96
7
)
6
5
8
o
0
0
Cu
3
0
00
2
00
00
0
0
-1
-15
8
2000
2500
3000
Speed (RPM)
Figure 47: Knock data, delayed with an approximate linear curve fit. Each point represents
the average over all knocking cycles at a fixed speed, load and spark timing.
Here is can be seen that it is possible to get a delay in auto-ignition of up to five crank angles
after peak pressure. Previously it was shown that five crank angles was the range needed to keep
the knock onset lines level, as in the experimental maps. The main result here is that the point at
which knock onset occurs during combustion depends on speed.
9. Revised Results
It was deemed appropriate to consider an alternate case in which the knock onset lines were less
dependent on speed, similar to the experimental results. In this new case the knock onset lines
are level, meaning that the bmep where knock onset is set to occur for a given fuel was
independent of the speed. In order to determine the relationship between load and ethanol
fraction the bmep where knock onset was set to occur was recorded for a variety of gasolineethanol blends. Since comparison between simulation and experiment were fairly good in the
1500 RPM to 2000 RPM range, the loads corresponding to 1750 RPM were used.
53
MBT Spark Timing
24
L
L
L
L
L
L
L.
L
-
2220
_
_
CR=9.2
CR=11.5
CR= 13.5
1816
UL!
14
12
10-
8 -6
4
0
5
10
15
20
25
30
35
40
45
50
Ethanol Percentage
Figure 48: BMEP as a function of ethanol fraction at MBT
Up to 5 degree retard
30
.
CR=9.2
CR=11.5'
CR=13.525
C', 20
M
-o
a-
151
r
10
5
50
5
10
15
20
25
30
35
40
45
50
Ethanol Percentage
Figure 49: BMEP v. ethanol fraction with 5 degree retard
54
Up to 10 degree retard
40
-
CR=9.2
CR=11.5
CR=13.5
3530
-0
25-
W 20 15
10
0
5
10
15
20
25
30
35
40
45
50
Ethanol Percentage
Figure 50: BMEP v. ethanol percentage at 10 degree retard
In the original performance maps the peak of the knock onset lines occurred at approximately
4000 RPM. As a result, for the majority of the performance map this new approach would lead
to higher ethanol usage.
9.1.
Ethanol Fraction
The generally expected trend is that with the new knock onset lines the required ethanol fraction
will be higher because the revised knock onset lines are horizontal and, for a majority of the
map, lower than the original map. This would require ethanol to be added at a lower point when
previously it would have been fine to keep using gasoline. This is most clearly seen in the US06
cycle. Here, ethanol fraction is up to 10 percentage points higher with the new map in each case.
The trend is reversed when dealing with some of the cases with UDDS and HWFET. Some of
the cases require lower ethanol percentages than the current model. This is also due to the fact
that the knock onset lines are horizontal. At low speeds points require ethanol in the original
map that did not in the new map, used with horizontal knock onset lines. As mentioned before,
some points require ethanol in one case, but not in the other.
55
UDDS Cycle
0.05
LL
0.045
0.04
o
0.035
-
9.2
11.5
13.5
0.03
0.0250
0.02S0.015
0.01 0.005
0-
0
1
2
3
4
5
6
7
8
9
10
Spark Retard
Figure 51: Comparison between original knock onset lines, (in bold) and new map with
horizontal knock onset lines (thin lines). The key result here is that the original case,
generally required more ethanol. This is due to the fact that the horizontal knock onset lines
require fewer ethanol at low loads and speeds.
56
HWFET Cycle
0.04:
0.035
C
0.03
0
9.2
11.5
13.5
-0.025
U-
-
0.02
C
CU 0. 015
0.01F
0.005
0
1
2
3
4
5
6
7
8
9
10
Spark Retard
Figure 52: Ethanol fraction for HWFET comparing original (thin lines) with new knock onset
constraints (bold lines)
L
0.35
USO6 Cycle
LL
LL
L
L
L-
0.3 C
0
0.25
9.2
__
.
11.5
13.5
--
0.2
U-
-
0.15
c
0.1
--
0.05
0
0
1
-----------
r
2
3
4
5
6
7
8
9
10
Spark Retard
Figure 53: Ethanol fraction for US06 comparing original (thin lines) with new knock onset
constraints (bold lines)
57
10. Findings and Conclusions
10.1. GT Power Modeling
GT Power was used to model experimental conditions in a turbocharged internal combustion
engine at speeds operating between 1500 RPM and 3000 RPM. This was done by calculating the
burn rate assuming poly-tropic expansion of burned species and poly-tropic compression of
unburned species. Several simulation parameters were modified in order to make simulation
agree with experiment. The best matches were found when combustion efficiency was set to
approximately 95%, the heat transfer model was scaled by 1 to 1.1 and it was assumed that 30%
of the available fuel - evaporation charge cooling was used to cool the walls and thus would not
be available to cool the mixture.
10.2. Comparison of Different Fuels
It was determined that at low loads the energy conversion efficiency is essentially independent of
the fuel used. At higher loads this is not the case because of limitations imposed by the type of
fuel and engine, i.e. knocking or peak pressure in the cylinder. When comparing different fuels,
the difference in efficiency was less than 1%. Therefore, it can be concluded that over the lower
two - thirds of the performance map, the energy conversion efficiency of the fuel is independent
of the fuel used.
10.3. Octane Number
GT Power was used to simulate experimental conditions in the cylinder. By matching the
pressure profile in GT Power with the experimental pressure profile it was possible to calculate
cylinder parameters that could not be measured in the current setup, such as mixture temperature.
By using the temperature and pressure of the unburned mixture from the beginning of
compression to knock onset, effective octane numbers could be generated. Octane numbers
found ranged from 91 for UTG91 gasoline to 111 for E25. The numbers found for these fuels
were, in some cases up to 10 octane numbers higher than those reported by other studies. These
were attributed to differences, engine to engine, which results from fuel sensitivities (RON MON). This amounts to two octane numbers. Another contributor to the difference is effect of
charge evaporative cooling which, for the fuels tested can be as high as ten octane numbers.
10.4. Higher Compression Ratio
In GT Power several performance maps were created at different compression ratios, assuming
that the fuels are interchangeable in the cylinder, other than knock constraints. This meant that it
was possible to run the engine at MBT indefinitely and mitigate knock by adding modest
amounts of ethanol within the cylinder. It was determined that an increase in compression ratio
from 9.2 to 13.5 leads to a relative increase in brake fuel conversion efficiency of 7 percent at
58
typical part load conditions; 30.9 to 33.2. This is supported by comparison with previous studies
of efficiency and compression ratio.
10.5. Spark Retard
In cases where it is desired to minimize the amount of ethanol, increasing spark retard is applied
first up to 5 or 10 crank angles after MBT timing if knock is predicted. Ethanol is added
afterwards once the maximum amount of spark retard is reached, depending of the severity of
knock. This would shift the knock onset lines upward, requiring smaller amounts of ethanol, but
at the same time it would decrease the efficiency of the engine at high loads. It was found in the
simulations that retarding spark from MBT timing by 5 and 10 crank angles after MBT timing
leads to relative efficiency drops of approximately 1.5 and 4 respectively.
10.6. Autonomie Results
Nine performance maps were generated corresponding to three compression ratios (9.2, 11.5 and
13.5) and three spark retard limits applied during light knock (MBT, retarded up to 5 crank angle
degrees and retarded up to 10 crank angle degrees). Each map was run using three drive cycles,
UDDS, HWFET and US06.
Effect of Compression Ratio: It was found that increasing compression ratio from 9.2 to 13.5
increases the miles per gallon of the drive cycle by 2 to 6 percent depending on the drive cycle.
The highest miles per gallon increase, in gasoline equivalent, was with the US06 cycle; about
5.7%. The smallest increase was with the UDDS cycle, 2.2%. This is lower than the engine
efficiency gain seen during part load.
Miles traveled per gallon of ethanol used was significantly higher than the gasoline value,
ranging from 141 miles per gallon of ethanol upward. Some of the low compression ratio cases
with spark retard required no ethanol because spark retard raised the knock onset lines above the
maximum torque of the drive cycle.
Effect of Spark Retard: Spark retard was introduced by dividing the performance map into
three regimes; MBT timing, retarded spark timing. MBT timing was used at loads low enough
that no ethanol was needed to prevent knock. Once knock onset was reached for UTG9 1, rather
than add ethanol, the spark timing was retarded up to a limit in crank angles after MBT timing.
Once the limit of spark retard was reached, ethanol was added and the spark was retarded at the
limit. This strategy increased the miles traveled per gallon of ethanol used by a factor of 2 to 6
since less ethanol was needed to suppress knock with retarded timing.
10.7. Dependence of Knock Onset on Speed
When comparing the knock onset lines in simulation to experimental results it was determined
that the experimental knock onset was less dependent on speed than the auto-ignition model
would suggest. It was determined that, experimentally, the BMEP where knock onset occurs
59
was essentially independent of speed. This was attributed to the fact that knock onset can occur
slightly later than at the peak pressure point in the course of combustion. By delaying the point
of knock onset it was determined that, in the 1500 RPM to 3000 RPM range, extending the
window of knock onset by 5 crank angle degrees after crank angle of peak pressure was more
than enough to create a level knock onset lines.
Experimentally it was determined that it is possible for knock onset to occur later than peak
pressure. By averaging over several knocking cases it was determined that it is possible for
knock onset to occur up to seven crank angle degrees after peak pressure. This is the likely
explanation of the discrepancy between the experiment and the simulation and makes horizontal
knock onset lines plausible.
10.8. Future Work
Knock Onset: The auto-ignition integral was used to solve for the octane number of different
fuels tested in the engine, and the octane number was then used to predict knock in cases beyond
what could be simulated. This was effective at low speeds by assuming knock onset occurs at
peak pressure, but at high speeds the model overestimated the load at which knock onset occurs.
A more accurate auto-ignition model must therefore be developed to take into account the fact
that knock onset can be delayed at higher speeds.
Engine Downsizing: Engine downsizing has the potential to yield efficiency gains, by taking
advantage in the relative changes of different efficiency loss mechanisms. When increasing the
compression ratio from 9.2 to 13.5, for example, the maximum BMEP of the engine increased by
1.41%. This means that, assuming the torque required out of the engine is the same, the engine
could be downsized by 1.41%. This would give about 1.41% less friction increasing the
efficiency of the overall engine.
Variables in the Simulation: The simulations only included constraints due to knock. In reality
other constraints limit the engine performance, such as: peak pressure in the cylinder, turbine exit
temperature, etc. Peak pressure at MBT reached almost 180 bars at the highest compression
ratio studied here. This was much higher than the limit of our experimental turbocharged engine
of 100 bars. A future study will include these additional limitations when generating
performance maps.
60
References
[1]
Heywood, J.B., Internal Combustion Engine Fundamentals, McGraw-Hill, 1988.
[21
Kasseris, E. and Heywood, J., "Charge Cooling Effects on Knock Limits in SI DI Engines Using
Gasoline/Ethanol Blends: Part 1-Quantifying Charge Cooling," SAE Technical Paper 2012-01-1275,
2012, doi:10.4271/2012-01-1275.
[31
Douaud, A. and Eyzat, P., "Four-Octane-Number Method for Predicting the Anti-Knock Behavior of
Fuels and Engines," SAE Technical Paper 780080, 1978, doi:10.4271/780080.
[41
Anderson, J., Leone, T., Shelby, M., Wallington, T. et al., "Octane Numbers of Ethanol-Gasoline
Blends: Measurements and Novel Estimation Method from Molar Composition," SAE Technical
Paper 2012-01-1274, 2012, doi:10.4271/2012-01-1274.
[51
Ayala, F., Gerty, M., and Heywood, J., "Effects of Combustion Phasing, Relative Air-fuel Ratio,
Compression Ratio, and Load on SI Engine Efficiency," SAE Technical Paper 2006-01-0229, 2006,
doi:10.4271/2006-01-0229.
[6] Grabner, P., EichIseder, H., and Eckhard, G., "Potential of E85 Direct Injection for Passenger Car
Application," SAE Technical Paper 2010-01-2086, 2010, doi:10.4271/2010-01-2086.
[7] Muranaka, S., Takagi, Y., and Ishida, T., "Factors Limiting the Improvement in Thermal Efficiency of
S. 1. Engine at Higher Compression Ratio," SAE Technical Paper 870548, 1987, doi:10.4271/870548.
[8] Jung, H., Leone, T., Shelby, M., Anderson, J. et al., "Fuel Economy and CO₂ Emissions of
Ethanol-Gasoline Blends in a Turbocharged DI Engine," SAE Int. J. Engines 6(1):422-434, 2013,
doi:10.4271/2013-01-1321.
[9] Mufioz, R., Han, Z., VanDerWege, B., and Yi, J., "Effect of Compression Ratio on Stratified-Charge
Direct- Injection Gasoline Combustion," SAE Technical Paper 2005-01-0100, 2005,
doi:10.4271/2005-01-0100.
[101 EPA, "Dynamometer Drive Schedules
I Testing & Measuring Emissions I US EPA," 6 February 2013.
<http://www.epa.gov/nvfel/testing/dynamometer.htm>
[11] Berry, I. M., "The Effects of Driving Style and Vehicle Performance on the Real-World Fuel
Consumption of U.S. Light-Duty Vehicles," Master of Science Thesis,Massachusetts Institute of
Technology, 2007.
[12] Toyota Website, "Features & Specs," Toyota Camry 2013, 2013, Web,
<http://www.toyota.com/camry/features.html#!/mpg/2514/2532/2546/2540>
61
62